Elsevier

Physics Letters A

Volume 146, Issue 4, 21 May 1990, Pages 204-208
Physics Letters A

Stochastic versus deterministic update in simulated annealing

https://doi.org/10.1016/0375-9601(90)90166-LGet rights and content

Abstract

We propose an algorithm which has several of the characteristics of simulated annealing but whose updating rule is deterministic. The results of the comparison between the two algorithms indicate that the stochasticity of the Metropolis updating in the simulated annealing algorithm does not play a major role in the search of near-optimal minima.

References (16)

  • T.J. Sejnowski et al.

    Physica D

    (1986)
  • S. Kirkpatrick et al.

    Science

    (1983)
  • E. Bonomi et al.

    SIAM Rev.

    (1984)
  • R.E. Randelman et al.

    J. Stat. Phys.

    (1986)
  • S. Patarnello et al.

    Europhys. Lett.

    (1987)
  • N. Metropolis et al.

    J. Chem. Phys.

    (1953)
  • C.M. Soukolis et al.

    Phys. Rev. B

    (1983)
  • G.S. Grest et al.

    Heidelberg Colloquium on Glassy dynamics

There are more references available in the full text version of this article.

Cited by (72)

  • A landscape-based analysis of fixed temperature and simulated annealing

    2023, European Journal of Operational Research
    Citation Excerpt :

    They conjecture that no monotone decreasing temperature sequence is optimal for a broader set of cases. They also consider a (deterministic) threshold random search, prove that there is an optimal sequence of threshold values, and state that probably in many situations there is an optimal deterministic threshold sequence that outperforms any random threshold sequence; incidentally, this can be considered the first study of deterministic variants of SA, later also called Threshold Acceptance (Dueck & Scheuer, 1990; Moscato & Fontanari, 1990). However, they add that “the practical implication of these likelihoods is clouded, since it is unclear how to efficiently find an optimal temperature sequence or deterministic threshold sequence for a problem instance” (Hajek & Sasaki, 1989), and thus SA is probably a good fallback solution.

  • Revisiting simulated annealing: A component-based analysis

    2019, Computers and Operations Research
    Citation Excerpt :

    For example, keeping the same temperature value throughout the whole execution turns SA into an algorithm known under several names, such as Metropolis Algorithm (Jerrum, 1992), Generalized Hill Climbing (Johnson and Jacobson, 2002), Static Simulated Annealing (Orosz and Jacobson, 2002), or simply fixed temperature schemes (Cohn and Fielding, 1999; Fielding, 2000). Replacing the probabilistic acceptance criterion with a deterministic one, it is possible to generate a new class of local search algorithms, such as the Threshold Acceptance (Dueck and Scheuer, 1990; Moscato and Fontanari, 1990), Great Deluge Algorithm and Record-to-Record Travel (Dueck, 1993), or the more recent Late Acceptance Hill Climbing (Burke and Bykov, 2012; 2017). All these variants are described in the next section.

View all citing articles on Scopus
1

Present address: Instituto de Fisica e Química de São Carlos, Universidade de São Paulo, 13560 São Carlos, SP, Brazil.

View full text